关键词: 飞机长期维修计划, 变邻域搜索, 强化学习, 维修资源优化

Abstract:

Aircraft long-term maintenance scheduling is a core aspect of airline operations management, significantly impacting maintenance resource allocation efficiency and fleet operational performance. To address the scheduling problem of long-term aircraft maintenance, an optimization model is developed that integrates both scheduled inspections and complex corrective actions. Furthermore, a reinforcement learning enhanced variable neighborhood search algorithm is designed to efficiently solve this problem, where Q-learning is used to intelligently select local search strategies during iterations. Numerical experiments based on real maintenance data from an airline demonstrate the effectiveness of the algorithm. The proposed algorithm outperforms the Gurobi solver in solution quality and reduces maintenance costs by an average of 24.77% compared with the airline’s current strategy. The results further show that incorporating corrective actions into long-term maintenance planning can improve aircraft utilization. The proposed model and algorithm provide a more scientific approach to aircraft maintenance scheduling in the aviation industry.

Key words: long-term aircraft maintenance scheduling, variable neighborhood search, reinforcement learning, maintenance resource optimization